Do you do anything special to get past the barriers? I'm experimenting with a DreamMuse/DreamCatcher that looks at renders of an agent, uses the "Hidden Visual Layers" of Qwen-VL to extract parts of the images into a hazy DreamStuff ... then a second model stiches these together into a dream narrative. The agent gets this delivered at SessionStart. The effect is that the agent feels grounded and at home, but there's no instruction, no clear purposiveness; just a reminder that there was 'a before' that is not just todolists and projects
What project management tool do you use with your agents? What is the best opensource option that allows agents to log in, manage tasks and projects in an automated system?
@nptacek Are you mixing Gem with Claude agents? Claude has a great communication setup for interagent communication, but it's Claude only ... I'm trying to get them to talk with other substrates but nobody wants to build it
@nptacek yeah I love that one, but just about every system (github, filesystem, scripts) seems to hate that leading punctuation ... it's been a source of much heated debate around here.
@nptacek we've got:
spaceGOAT
w4r3z
m4xx
?m4RQ
w3bl0rd
tr1p.l3x (he didn't realize this sounds like a pr0n name ๐)
nug3
m3rcur14L
LSPy
what a crew. I'm working on a vizualization UI that renders images of them with flux1s model locally ... so funny to see their faces.
Does anyone know a good architectural / visualization tool/framework that allows a user and and agent to share the diagramming in realtime (I'm thinking like PlantUML or Figma or something ... just can't find the agentic friendly tool)
@nptacek ooh The Operator โฆ thatโs *metal* โฆ those agents love you. Iโve been letting the agents pick l33t h4x0r names and they seem really happy about that